What this column is, and the five rules it runs on

I spent eleven years on a fixed-income desk, most of it in relative value and the last two in a seat that mostly involved explaining relative value to people who did not want it explained. I have now been writing about markets for longer than I traded them, which is either a warning or a qualification.

Rule one: one column a week

Thursday, most weeks. If there is nothing worth saying there is no column, and that has happened four times. A weekly slot that must be filled produces the kind of writing that fills a weekly slot.

Rule two: no calls, no model portfolio

I do not publish price targets, a recommended allocation, or a list of positions to copy. Partly because I would be bad at it and partly because a published call changes what the column is for. Once you have a scoreboard, everything you write is defending the scoreboard.

Rule three: show the number or do not make the claim

Every assertion with a magnitude in it gets a table. If I cannot produce the table, the sentence comes out. This is the rule that kills the most drafts.

Rule four: guests have to have done the job

Guest columns come from people with operating experience in the thing they are writing about — a regulatory affairs job, a municipal desk, an audit seat. Not from strategists, and not from anyone who wants the column to link to something they are selling. Guests disclose in their own words at the foot of their piece.

Rule five: retire what has gone stale

Commentary about a specific market condition stops being commentary and starts being misinformation once the condition ends. Columns tied to a particular rate environment come down when that environment does, rather than sitting in an archive being found by someone searching next year. The archive says what has been retired.

How it is made

Written in a text editor, published as plain files. The data in the tables comes from primary sources — filings, tape data, official calendars — and where I have processed it myself the processing is described in the column. Where a figure is a rough estimate it says so with the word “roughly”, which appears more often than I would like.